In 2025, Anthropic stood nearly alone among frontier AI labs in publicly endorsing landmark AI transparency laws in California and New York. Most of Silicon Valley fought those bills. Today, Anthropic's head of state government relations says those laws are already outdated, and the company is pressing state legislatures to pass something far more aggressive. For founders building on top of AI platforms, this shift matters. The rules that govern how AI models are built, audited, and deployed are being written right now, and one of the companies writing them is also the one selling you the models.
The Strategy Behind Pushing Tougher Rules
Cesar Fernandez, Anthropic's head of US state and local government relations, told WIRED that transparency and self-reporting are no longer enough for the most powerful AI systems. The company has endorsed an Illinois measure requiring third-party auditor evaluations of AI safety processes and a Massachusetts policy that would let the state attorney general seek injunctive relief against non-compliant labs. These are not small changes. Third-party auditing requirements create real operational burdens. They mean publishing detailed safety documentation, submitting to external review, and potentially pausing deployments if the auditor finds problems. For a startup operating on thin margins and fast release cycles, that is a genuine bottleneck. Anthropic says these rules should only apply to large AI model developers, defined as companies spending hundreds of millions on AI with over $500 million in annual revenue. In practice, that threshold excludes most startups, but the gap is narrowing. Safe Superintelligence, Thinking Machines Lab, and Mistral have each raised billions in capital, and they are approaching that revenue bar fast.
Regulatory Capture or Responsible Governance?
Former White House AI czar David Sacks has publicly accused Anthropic of running a sophisticated regulatory capture strategy. His argument is straightforward: by pushing for regulations that are expensive to comply with, Anthropic locks in its own market position while smaller competitors struggle under the compliance load. The accusation is not new, but it is gaining traction as Anthropic's influence over state-level AI policy becomes more visible. Fernandez denies the charge, pointing out that Anthropic only supports bills targeting large model developers. But the line between large and small is blurring quickly in an industry where a single training run can cost half a billion dollars. Anthropic's valuation has crossed the trillion-dollar mark. Whether or not its motives are pure, its policy positions shape the regulatory environment that every AI startup must navigate. For founders, the question is not whether Anthropic is being sincere. The question is whether the rules being written now will create moats that only incumbents can cross.
Where Anthropic Draws Its Own Lines
One telling detail in the WIRED interview reveals the limits of Anthropic's pro-regulation stance. While the company supports state-level auditing and transparency requirements, it believes the power to block unsafe AI model deployments should remain with the federal government, not state lawmakers. This distinction is strategic. Federal preemption would create a single national standard, which is easier for companies like Anthropic to comply with than a patchwork of 50 state laws. It also consolidates regulatory power at a level where Anthropic already maintains a substantial lobbying presence. The Trump administration recently forced Anthropic to take its two most powerful models offline for foreign nationals under an export control directive. Anthropic was not thrilled about that. The company argued that blocking model deployments should happen through a fair, transparent evaluation process, not executive action. Translation: Anthropic wants regulation, but only the kind it can predict, plan for, and possibly influence.
What This Means for Founders and Builders
Three dynamics are worth watching. First, the gap between what regulators want and what users want is growing. Voters consistently rank job displacement and data center community impacts above catastrophic AI risks, yet frontier labs have focused their political efforts almost entirely on existential safety. Second, the open-weight model debate is increasingly being fought through regulation rather than technology. Anthropic's letter to the US government accusing Alibaba of a distillation attack is widely seen as a move to ban Chinese open-weight models, which thousands of US businesses rely on. If those models get restricted, those businesses will need alternatives, and Anthropic's enterprise offerings will be waiting. Third, the state-level push is creating a compliance environment that rewards incumbents. Whether you believe Anthropic is acting in good faith or not, the outcome is the same: the rules being written today are raising the cost of entry for AI builders. Founders should pay close attention to their state's AI legislation, build compliance processes early, and understand that the regulatory game is now being played in state capitols, not just Washington.